Abstract 4008: Transcriptomic signatures in esophageal adenocarcinoma define distinct subtypes with therapeutic relevance

Abstract Background & Aims: Esophageal cancer is the fifth most common solid cancer globally, and esophageal adenocarcinoma (EAC) is the predominant histological subtype in the western world. Patients often present at an advanced stage and overall-5-year survival rates are less than 15%. In many patients, the response to neoadjuvant therapy is encouraging at first, but most will develop metastatic disease several years later. Recently, we observed plasticity along the epithelial-to-mesenchymal (EMT) axis in EAC tumor cells upon therapeutic pressure. The last decade has seen the discovery of static molecular subtypes in cancers and a poor-prognosis mesenchymal subtype has been reported in multiple gastrointestinal cancers, but not in EAC. Here, we set out to identify transcriptomic subtypes and the frequency of their occurrence during EAC progression, and exposure to therapy in a clinical setting. Methods: A cohort comprising of 174 esophageal cancer cases (n=100 pre-treatment biopsies, n=46 neoadjuvantly-treated resection specimens, and n=28 metastatic biopsies) with 12 esophageal cancer cell lines (8 primary and 4 ATCC cell lines) was RNA-sequenced. To circumvent non-tumor signals confounding the analysis, we leveraged non-negative matrix factorization (NMF) as a virtual microdissection tool. Patients were subtyped by consensus clustering the top 50 exemplar genes per tumor-cell intrinsic signature. Next, the subtypes were validated in two independent cohorts and clinical correlates were analyzed. In addition, we perturbed subtype-specific regulators to functionally assess the existence of the subtypes. Results: Here, we identified 7 biological signatures involved in primary and metastatic EAC. Two of these were unambiguously tumor-intrinsic signals. All seven signatures associated with tumor cellularity scores estimated by an experienced pathologist confirming the reliability of NMF as a virtual microdissection tool. Upon clustering two EAC subtypes emerged: Intestinal-like (IL) and Mesenchymal-like (ML). ML-subtyped patients were observed with increased frequency after neoadjuvant treatment, and following metastatic dissemination. This association was validated in an independent cohort. Lastly, we identified HMGA2 as a key transcription factor for the ML-subtype. Genetic perturbation of HMGA2 induced an IL-associated phenotype. Conclusion: Our study demonstrate for the first time from clinical transcriptomics data that EAC cells can exist in distinct cell states. In addition, the abundance of either cell states, that constitute to tumor subtype, alter when exposed to chemoradiation or following metastatic spread. These subtypes offer a valuable method to design more informed personalized treatment approaches for EAC patients; for existing and future subtype-directed treatments. Citation Format: Mark P.G. Dings, Amber P. Zalm, Marjolein F. Lansbergen, César Oyarce, Sybren L. Meijer, Cynthia Waasdorp, Jan Paul Medema, Hanneke van Laarhoven, Maarten F. Bijlsma. Transcriptomic signatures in esophageal adenocarcinoma define distinct subtypes with therapeutic relevance [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2022; 2022 Apr 8-13. Philadelphia (PA): AACR; Cancer Res 2022;82(12_Suppl):Abstract nr 4008.